fsck-objects: refactor checking for connectivity
[git/dscho.git] / sha1_name.c
blob9dfb3ac57403793e6c6f78f2f75fb61c0a56c4bc
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tree-walk.h"
7 #include "refs.h"
9 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
11 struct alternate_object_database *alt;
12 char hex[40];
13 int found = 0;
14 static struct alternate_object_database *fakeent;
16 if (!fakeent) {
17 const char *objdir = get_object_directory();
18 int objdir_len = strlen(objdir);
19 int entlen = objdir_len + 43;
20 fakeent = xmalloc(sizeof(*fakeent) + entlen);
21 memcpy(fakeent->base, objdir, objdir_len);
22 fakeent->name = fakeent->base + objdir_len + 1;
23 fakeent->name[-1] = '/';
25 fakeent->next = alt_odb_list;
27 sprintf(hex, "%.2s", name);
28 for (alt = fakeent; alt && found < 2; alt = alt->next) {
29 struct dirent *de;
30 DIR *dir;
31 sprintf(alt->name, "%.2s/", name);
32 dir = opendir(alt->base);
33 if (!dir)
34 continue;
35 while ((de = readdir(dir)) != NULL) {
36 if (strlen(de->d_name) != 38)
37 continue;
38 if (memcmp(de->d_name, name + 2, len - 2))
39 continue;
40 if (!found) {
41 memcpy(hex + 2, de->d_name, 38);
42 found++;
44 else if (memcmp(hex + 2, de->d_name, 38)) {
45 found = 2;
46 break;
49 closedir(dir);
51 if (found == 1)
52 return get_sha1_hex(hex, sha1) == 0;
53 return found;
56 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
58 do {
59 if (*a != *b)
60 return 0;
61 a++;
62 b++;
63 len -= 2;
64 } while (len > 1);
65 if (len)
66 if ((*a ^ *b) & 0xf0)
67 return 0;
68 return 1;
71 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
73 struct packed_git *p;
74 unsigned char found_sha1[20];
75 int found = 0;
77 prepare_packed_git();
78 for (p = packed_git; p && found < 2; p = p->next) {
79 unsigned num = num_packed_objects(p);
80 unsigned first = 0, last = num;
81 while (first < last) {
82 unsigned mid = (first + last) / 2;
83 unsigned char now[20];
84 int cmp;
86 nth_packed_object_sha1(p, mid, now);
87 cmp = hashcmp(match, now);
88 if (!cmp) {
89 first = mid;
90 break;
92 if (cmp > 0) {
93 first = mid+1;
94 continue;
96 last = mid;
98 if (first < num) {
99 unsigned char now[20], next[20];
100 nth_packed_object_sha1(p, first, now);
101 if (match_sha(len, match, now)) {
102 if (nth_packed_object_sha1(p, first+1, next) ||
103 !match_sha(len, match, next)) {
104 /* unique within this pack */
105 if (!found) {
106 hashcpy(found_sha1, now);
107 found++;
109 else if (hashcmp(found_sha1, now)) {
110 found = 2;
111 break;
114 else {
115 /* not even unique within this pack */
116 found = 2;
117 break;
122 if (found == 1)
123 hashcpy(sha1, found_sha1);
124 return found;
127 #define SHORT_NAME_NOT_FOUND (-1)
128 #define SHORT_NAME_AMBIGUOUS (-2)
130 static int find_unique_short_object(int len, char *canonical,
131 unsigned char *res, unsigned char *sha1)
133 int has_unpacked, has_packed;
134 unsigned char unpacked_sha1[20], packed_sha1[20];
136 has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
137 has_packed = find_short_packed_object(len, res, packed_sha1);
138 if (!has_unpacked && !has_packed)
139 return SHORT_NAME_NOT_FOUND;
140 if (1 < has_unpacked || 1 < has_packed)
141 return SHORT_NAME_AMBIGUOUS;
142 if (has_unpacked != has_packed) {
143 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
144 return 0;
146 /* Both have unique ones -- do they match? */
147 if (hashcmp(packed_sha1, unpacked_sha1))
148 return SHORT_NAME_AMBIGUOUS;
149 hashcpy(sha1, packed_sha1);
150 return 0;
153 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
154 int quietly)
156 int i, status;
157 char canonical[40];
158 unsigned char res[20];
160 if (len < MINIMUM_ABBREV || len > 40)
161 return -1;
162 hashclr(res);
163 memset(canonical, 'x', 40);
164 for (i = 0; i < len ;i++) {
165 unsigned char c = name[i];
166 unsigned char val;
167 if (c >= '0' && c <= '9')
168 val = c - '0';
169 else if (c >= 'a' && c <= 'f')
170 val = c - 'a' + 10;
171 else if (c >= 'A' && c <='F') {
172 val = c - 'A' + 10;
173 c -= 'A' - 'a';
175 else
176 return -1;
177 canonical[i] = c;
178 if (!(i & 1))
179 val <<= 4;
180 res[i >> 1] |= val;
183 status = find_unique_short_object(i, canonical, res, sha1);
184 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
185 return error("short SHA1 %.*s is ambiguous.", len, canonical);
186 return status;
189 const char *find_unique_abbrev(const unsigned char *sha1, int len)
191 int status, is_null;
192 static char hex[41];
194 is_null = is_null_sha1(sha1);
195 memcpy(hex, sha1_to_hex(sha1), 40);
196 if (len == 40 || !len)
197 return hex;
198 while (len < 40) {
199 unsigned char sha1_ret[20];
200 status = get_short_sha1(hex, len, sha1_ret, 1);
201 if (!status ||
202 (is_null && status != SHORT_NAME_AMBIGUOUS)) {
203 hex[len] = 0;
204 return hex;
206 if (status != SHORT_NAME_AMBIGUOUS)
207 return NULL;
208 len++;
210 return NULL;
213 static int ambiguous_path(const char *path, int len)
215 int slash = 1;
216 int cnt;
218 for (cnt = 0; cnt < len; cnt++) {
219 switch (*path++) {
220 case '\0':
221 break;
222 case '/':
223 if (slash)
224 break;
225 slash = 1;
226 continue;
227 case '.':
228 continue;
229 default:
230 slash = 0;
231 continue;
233 break;
235 return slash;
238 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
240 static const char *fmt[] = {
241 "%.*s",
242 "refs/%.*s",
243 "refs/tags/%.*s",
244 "refs/heads/%.*s",
245 "refs/remotes/%.*s",
246 "refs/remotes/%.*s/HEAD",
247 NULL
249 const char **p, *r;
250 int refs_found = 0;
252 *ref = NULL;
253 for (p = fmt; *p; p++) {
254 unsigned char sha1_from_ref[20];
255 unsigned char *this_result;
257 this_result = refs_found ? sha1_from_ref : sha1;
258 r = resolve_ref(mkpath(*p, len, str), this_result, 1, NULL);
259 if (r) {
260 if (!refs_found++)
261 *ref = xstrdup(r);
262 if (!warn_ambiguous_refs)
263 break;
266 return refs_found;
269 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
271 static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
272 char *real_ref = NULL;
273 int refs_found = 0;
274 int at, reflog_len;
276 if (len == 40 && !get_sha1_hex(str, sha1))
277 return 0;
279 /* basic@{time or number} format to query ref-log */
280 reflog_len = at = 0;
281 if (str[len-1] == '}') {
282 for (at = 1; at < len - 1; at++) {
283 if (str[at] == '@' && str[at+1] == '{') {
284 reflog_len = (len-1) - (at+2);
285 len = at;
286 break;
291 /* Accept only unambiguous ref paths. */
292 if (ambiguous_path(str, len))
293 return -1;
295 refs_found = dwim_ref(str, len, sha1, &real_ref);
297 if (!refs_found)
298 return -1;
300 if (warn_ambiguous_refs && refs_found > 1)
301 fprintf(stderr, warning, len, str);
303 if (reflog_len) {
304 /* Is it asking for N-th entry, or approxidate? */
305 int nth, i;
306 unsigned long at_time;
307 unsigned long co_time;
308 int co_tz, co_cnt;
310 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
311 char ch = str[at+2+i];
312 if ('0' <= ch && ch <= '9')
313 nth = nth * 10 + ch - '0';
314 else
315 nth = -1;
317 if (0 <= nth)
318 at_time = 0;
319 else
320 at_time = approxidate(str + at + 2);
321 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
322 &co_time, &co_tz, &co_cnt)) {
323 if (at_time)
324 fprintf(stderr,
325 "warning: Log for '%.*s' only goes "
326 "back to %s.\n", len, str,
327 show_rfc2822_date(co_time, co_tz));
328 else
329 fprintf(stderr,
330 "warning: Log for '%.*s' only has "
331 "%d entries.\n", len, str, co_cnt);
335 free(real_ref);
336 return 0;
339 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
341 static int get_parent(const char *name, int len,
342 unsigned char *result, int idx)
344 unsigned char sha1[20];
345 int ret = get_sha1_1(name, len, sha1);
346 struct commit *commit;
347 struct commit_list *p;
349 if (ret)
350 return ret;
351 commit = lookup_commit_reference(sha1);
352 if (!commit)
353 return -1;
354 if (parse_commit(commit))
355 return -1;
356 if (!idx) {
357 hashcpy(result, commit->object.sha1);
358 return 0;
360 p = commit->parents;
361 while (p) {
362 if (!--idx) {
363 hashcpy(result, p->item->object.sha1);
364 return 0;
366 p = p->next;
368 return -1;
371 static int get_nth_ancestor(const char *name, int len,
372 unsigned char *result, int generation)
374 unsigned char sha1[20];
375 int ret = get_sha1_1(name, len, sha1);
376 if (ret)
377 return ret;
379 while (generation--) {
380 struct commit *commit = lookup_commit_reference(sha1);
382 if (!commit || parse_commit(commit) || !commit->parents)
383 return -1;
384 hashcpy(sha1, commit->parents->item->object.sha1);
386 hashcpy(result, sha1);
387 return 0;
390 static int peel_onion(const char *name, int len, unsigned char *sha1)
392 unsigned char outer[20];
393 const char *sp;
394 unsigned int expected_type = 0;
395 struct object *o;
398 * "ref^{type}" dereferences ref repeatedly until you cannot
399 * dereference anymore, or you get an object of given type,
400 * whichever comes first. "ref^{}" means just dereference
401 * tags until you get a non-tag. "ref^0" is a shorthand for
402 * "ref^{commit}". "commit^{tree}" could be used to find the
403 * top-level tree of the given commit.
405 if (len < 4 || name[len-1] != '}')
406 return -1;
408 for (sp = name + len - 1; name <= sp; sp--) {
409 int ch = *sp;
410 if (ch == '{' && name < sp && sp[-1] == '^')
411 break;
413 if (sp <= name)
414 return -1;
416 sp++; /* beginning of type name, or closing brace for empty */
417 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
418 expected_type = OBJ_COMMIT;
419 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
420 expected_type = OBJ_TREE;
421 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
422 expected_type = OBJ_BLOB;
423 else if (sp[0] == '}')
424 expected_type = OBJ_NONE;
425 else
426 return -1;
428 if (get_sha1_1(name, sp - name - 2, outer))
429 return -1;
431 o = parse_object(outer);
432 if (!o)
433 return -1;
434 if (!expected_type) {
435 o = deref_tag(o, name, sp - name - 2);
436 if (!o || (!o->parsed && !parse_object(o->sha1)))
437 return -1;
438 hashcpy(sha1, o->sha1);
440 else {
441 /* At this point, the syntax look correct, so
442 * if we do not get the needed object, we should
443 * barf.
446 while (1) {
447 if (!o || (!o->parsed && !parse_object(o->sha1)))
448 return -1;
449 if (o->type == expected_type) {
450 hashcpy(sha1, o->sha1);
451 return 0;
453 if (o->type == OBJ_TAG)
454 o = ((struct tag*) o)->tagged;
455 else if (o->type == OBJ_COMMIT)
456 o = &(((struct commit *) o)->tree->object);
457 else
458 return error("%.*s: expected %s type, but the object dereferences to %s type",
459 len, name, typename(expected_type),
460 typename(o->type));
461 if (!o->parsed)
462 parse_object(o->sha1);
465 return 0;
468 static int get_describe_name(const char *name, int len, unsigned char *sha1)
470 const char *cp;
472 for (cp = name + len - 1; name + 2 <= cp; cp--) {
473 char ch = *cp;
474 if (hexval(ch) & ~0377) {
475 /* We must be looking at g in "SOMETHING-g"
476 * for it to be describe output.
478 if (ch == 'g' && cp[-1] == '-') {
479 cp++;
480 len -= cp - name;
481 return get_short_sha1(cp, len, sha1, 1);
485 return -1;
488 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
490 int ret, has_suffix;
491 const char *cp;
493 /* "name~3" is "name^^^",
494 * "name~" and "name~0" are name -- not "name^0"!
495 * "name^" is not "name^0"; it is "name^1".
497 has_suffix = 0;
498 for (cp = name + len - 1; name <= cp; cp--) {
499 int ch = *cp;
500 if ('0' <= ch && ch <= '9')
501 continue;
502 if (ch == '~' || ch == '^')
503 has_suffix = ch;
504 break;
507 if (has_suffix) {
508 int num = 0;
509 int len1 = cp - name;
510 cp++;
511 while (cp < name + len)
512 num = num * 10 + *cp++ - '0';
513 if (has_suffix == '^') {
514 if (!num && len1 == len - 1)
515 num = 1;
516 return get_parent(name, len1, sha1, num);
518 /* else if (has_suffix == '~') -- goes without saying */
519 return get_nth_ancestor(name, len1, sha1, num);
522 ret = peel_onion(name, len, sha1);
523 if (!ret)
524 return 0;
526 ret = get_sha1_basic(name, len, sha1);
527 if (!ret)
528 return 0;
530 /* It could be describe output that is "SOMETHING-gXXXX" */
531 ret = get_describe_name(name, len, sha1);
532 if (!ret)
533 return 0;
535 return get_short_sha1(name, len, sha1, 0);
539 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
540 * notably "xyz^" for "parent of xyz"
542 int get_sha1(const char *name, unsigned char *sha1)
544 int ret, bracket_depth;
545 unsigned unused;
546 int namelen = strlen(name);
547 const char *cp;
549 prepare_alt_odb();
550 ret = get_sha1_1(name, namelen, sha1);
551 if (!ret)
552 return ret;
553 /* sha1:path --> object name of path in ent sha1
554 * :path -> object name of path in index
555 * :[0-3]:path -> object name of path in index at stage
557 if (name[0] == ':') {
558 int stage = 0;
559 struct cache_entry *ce;
560 int pos;
561 if (namelen < 3 ||
562 name[2] != ':' ||
563 name[1] < '0' || '3' < name[1])
564 cp = name + 1;
565 else {
566 stage = name[1] - '0';
567 cp = name + 3;
569 namelen = namelen - (cp - name);
570 if (!active_cache)
571 read_cache();
572 if (active_nr < 0)
573 return -1;
574 pos = cache_name_pos(cp, namelen);
575 if (pos < 0)
576 pos = -pos - 1;
577 while (pos < active_nr) {
578 ce = active_cache[pos];
579 if (ce_namelen(ce) != namelen ||
580 memcmp(ce->name, cp, namelen))
581 break;
582 if (ce_stage(ce) == stage) {
583 hashcpy(sha1, ce->sha1);
584 return 0;
586 pos++;
588 return -1;
590 for (cp = name, bracket_depth = 0; *cp; cp++) {
591 if (*cp == '{')
592 bracket_depth++;
593 else if (bracket_depth && *cp == '}')
594 bracket_depth--;
595 else if (!bracket_depth && *cp == ':')
596 break;
598 if (*cp == ':') {
599 unsigned char tree_sha1[20];
600 if (!get_sha1_1(name, cp-name, tree_sha1))
601 return get_tree_entry(tree_sha1, cp+1, sha1,
602 &unused);
604 return ret;